RECONSIDERING TAX EXPENDITURE ESTIMATION: CHALLENGES AND REFORMS

by

Rosanne Altshuler Rutgers University

altshule@rci.rutgers.edu

and

Robert D. Dietz National Association of Home Builders

rdietz@nahb.com

This draft: June 2, 2008

Abstract We examine the measurement of tax expenditures and offer recommendations aimed at

improving their value to analysts and policymakers. We use calculations from NBER’s TAXSIM to illustrate some of the problems with the current methodology for estimating tax expenditures. Unlike most previous work on the topic, we focus on how features of the current tax system complicate and compromise the value of information provided by the tax expenditure budget. Our recommendations for reform include presenting revenue estimates for major tax expenditures, reporting some negative tax expenditures, grouping expenditures by conceptually-linked categories, and establishing an appendix for tax expenditure estimates of permanent versions of expiring provisions and AMT interaction effects.

This paper was prepared for presentation at the NBER Conference, “Incentive and Distributional Consequences of Tax Expenditures,” held in Bonita Springs, FL on March 27-29, 2008. We thank Daniel Feenberg for providing invaluable help with TAXSIM. We thank Thomas Barthold, Edward Kleinbard, and conference attendees for valuable comments. We also thank Patrick Driessen for helpful comments on an earlier draft and Gary Koenig and Tim Dowd for useful suggestions. Nothing in this paper should be construed as reflecting the views of the National Association of Home Builders or its members.

INTRODUCTION

To most tax policy analysts, the term “tax expenditure” means a special provision of the

tax system that results in reduced tax liability for certain subsets of taxpayers. Moreover, for

many in the tax policy community, the term suggests tax breaks for limited constituencies that

result in a narrow tax base and higher marginal tax rates. Some are blunter: tax expenditures are

loopholes that need to be closed. We leave it to other authors to examine the legitimacy of tax

expenditure provisions from tax policy or economic efficiency perspectives. Instead, the present

analysis examines the measurement of tax expenditures and offers recommendations to improve

their value to analysts and policymakers.

The technical definition of tax expenditure is found in the Congressional Budget and

Impoundment Act of 1974 (“Budget Act”), which established the modern Congressional budget-

making process.1 The Budget Act requires annual publication of a list of tax expenditures in

order to improve the transparency of the federal government budget and account for proxy

spending programs governed by the nation’s tax code. The Budget Act formally defines a tax

expenditure as:

“Revenue losses attributable to provisions of Federal income tax laws which allow a special exclusion, exemption, or deduction from gross income or which provide a special credit, a preferential rate of tax, or a deferral of tax liability.”

As is well known, this definition is open to considerable interpretation. The legislative

history for the Budget Act indicates that tax expenditures are to be determined in reference to

“normal income tax law.” While not codified, legislative history can provide guidance for

implementing the law. However, determining what is “normal” is an open question and almost

certainly a normative exercise.

1 Public Law 93-344.

Interestingly, by the time the Budget Act was enacted both the Congressional Joint

Committee on Taxation (JCT) and the Department of the Treasury Office of Tax Analysis (OTA)

had been producing tax expenditure analysis for years. The term “tax expenditure” is attributed

to Assistant Secretary of the Treasury (Tax Policy) Stanley Surrey.2 In 1967 Surrey instructed

his staff to compile a list of preferences and concessions in the income tax that were similar to

expenditure programs. While Surrey’s stated motivation was to improve the budget process, he

was also interested in drawing attention to subsidies in the tax code in hopes of building

momentum for base-broadening tax reform.

Surrey’s efforts resulted in Treasury’s first tax expenditure budget report in 1968.

Classifying certain provisions of the Internal Revenue Code as deviations from a comprehensive,

progressive income tax, the Department of the Treasury reported tax expenditures for the

individual and corporate income tax systems, a process that has continued to the present.3 While

the first tax expenditure lists from Treasury were not included in the official budget, the Budget

Act mandated that tax expenditures be reported as part of the Administration budget. The Budget

Act required no distributional or other economic analysis beyond an estimate of the magnitude of

the cost in terms of lost revenue of each listed tax expenditure.

Reflecting reconsiderations in the concept and presentation of tax expenditures, the

Treasury presentation has evolved over time. Starting with the FY 1983 Budget, Treasury

introduced an additional tax baseline to classify and estimate tax expenditures. The new baseline,

called the “reference” tax baseline, is closer to present law than the normal tax baseline and

results in a more narrow definition of tax expenditures. Beginning with the FY 1984 Budget and 2 As Shaviro (2003) notes, the tax expenditure concept seems to have been invented twice, or at the very least, earlier elsewhere. As early as 1954, writers in Germany had noticed the equivalence between special tax deductions, credits, and other allowances and government subsidies. As a result, in 1959, the German government began reporting on subsidies in the federal budget, including those supplied through the tax system. Budgetary reports on tax subsidies that were classified as indirect spending were being made regularly by 1967. 3 Until 2003, Treasury also included a list of tax expenditures against a transfer-tax baseline (estate and gift taxes). Tax expenditures for other taxes, such as excise taxes, are not reported. As always, there are limited exceptions. For example, both JCT and Treasury have reported the reduction in excise tax receipts that result from the alcohol fuel and biodiesel credits.

2

until recently, Treasury presented outlay equivalent estimates for tax expenditures in addition to

the traditional revenue-based estimates.4 The goal of this approach is to provide estimates that

more closely correspond to estimates of direct outlay programs.5 Starting with the FY 1995

Budget, Treasury has reported present-value calculations of tax expenditures that involve either

the deferral of tax payments into future periods or other long-term effects.

Recently, the Administration has expressed concern with the arbitrary tax base used to

calculate tax expenditures. The FY 2002 Budget states that “(B)ecause of the breadth of this

arbitrary tax base, the Administration believes that the concept of ‘tax expenditure’ is uncertain

(OMB 2002, p. 61).” The FY 2003 Budget describes a Treasury Department effort to reconsider

and revise the tax expenditure presentation and notes that “(D)ue, in part, to the degree of

arbitrariness in the tax expenditure baseline, the Administration believes the meaningfulness of

tax expenditure estimates is uncertain (OMB 2003, p.95).” To address these concerns, the

Administration began estimating tax expenditure estimates against a comprehensive income tax

baseline and a consumption tax baseline in its FY 2004 Budget. In addition, the FY 2004 Budget

introduced a number of innovations including changes in the accelerated depreciation baseline,

the inclusion of or identification of negative tax expenditures, an estimate of the tax expenditure

for net imputed rent for owner-occupied housing, and an estimate of the (negative) tax

expenditure for the double-taxation of corporate income against a comprehensive income tax.

4 The Administration explains in its FY 2008 Budget that outlay equivalents are no longer included “because they were often the same as the normal tax expenditure estimates, and the criteria for applying the concepts as to when they should differ were often judgmental and hard to apply with consistency across time and across tax expenditure items (OMB 2008, p. 286).” 5 While the linkage of tax expenditures to direct spending programs has conceptual merit, there are numerous analytical complications that hinder reporting comparable numbers for these two classes of government fiscal policy. Although government accounting is not perfect, auditors and policymakers know with reasonable confidence what resources the federal government dedicates to various spending programs. The same cannot be said of all tax expenditures because of various information asymmetries. The government has access to its spending records. It does not necessarily have access to tax expenditure data. To take just one example, income exclusions that involve no information reporting (e.g. enterprise zone capital gain provisions) cannot be estimated with IRS data. Thus, such tax expenditure estimates rely on the educated guesses of government economists rather than the tabulation of reports by accountants.

3

The JCT began to estimate tax expenditures in 1972. In contrast to the Treasury analysis,

the JCT presentation had remained relatively constant, usefully allowing analysts to make

intertemporal comparisons. However, recently JCT announced planned changes to its methods of

presentation and measurement of tax expenditures (see JCT 2008). The most significant change

involves a revision of the classification of tax expenditures. JCT announced it will identify tax

expenditures as either “Tax Subsidies” (tax law provisions that are deliberately inconsistent with

an identifiable general rule of present law) or “Tax-Induced Structural Distortions” (structural

elements that materially affect economic decisions in a manner that imposes substantial economic

efficiency costs).6 Further, JCT has announced that it will now include negative tax expenditures

in its analysis. Finally, the revision will use a reference baseline for estimation purposes. This

last change reflects an acknowledgement of the limits of defining tax expenditures relative to a

normal income tax baseline. The JCT has indicated that it will publish more details of its

proposed changes later in 2008.

The evolution of the Treasury and JCT tax expenditure reports reflect, at least in part, the

concerns of analysts writing on the subject from within and outside of government agencies (both

in the United States and abroad).7 Legal scholars and economists have wrestled with the tax

expenditure concept since Surrey developed the idea.8 A growing literature debates the

usefulness of tax expenditure analysis for expenditure control, social and economic policy, and

tax reform.9 And recommendations to improve tax expenditure reporting by providing more

6 JCT will further divide tax expenditures labeled as Tax Subsidies as Tax Transfers (payments made to taxpayers with regard to tax liability; e.g. child tax credit), Social Spending (policies intended to subsidize or induce behavior; e.g. deduction for charitable giving), or Business Synthetic Spending (provisions intended to subsidize or induce behavior directly related to the production of business or investment income, excluding Tax Subsidies related to labor supply; e.g. section 199 deduction). 7 The literature on tax expenditures is too extensive to list here. For review of tax expenditure practices in other countries see Polackova et. al. (2004). 8 See Bittker (1969), Andrews (1972), Surrey (1973), Fiekowsky (1980), Surrey and McDaniel (1985), Thuronyi (1988) and Shaviro (2003), for example. 9 See, for example, Toder (2002), Burman (2003), and Toder (2005).

4

detail have recently been put forward by the Century Foundation Working Group on Tax

Expenditures (Century Foundation Press 2002).10

In this paper, we address the methodology of tax expenditure estimation. We have two

goals: to illustrate some of the problems with the current methodology and to formulate easily

implementable recommendations for reporting reform. Unlike most previous work on the topic,

we focus on features of the current tax system that were not in place when Surrey introduced the

expenditure concept, such as the alternative minimum tax (AMT) and sunset rules, and how these

features complicate and compromise the value of information provided by the tax expenditure

reporting. We begin in the next section with a discussion of the issues that arise in defining the

“normal” income tax structure. We review how this definition has been implemented by the JCT

and Treasury and illustrate some of the consequences of the choice of the normal tax structure

using estimates from NBER’s TAXSIM model. The third section reviews the basics of tax

expenditure estimation, explores how “scoring” conventions affect the information provided by

estimates, and considers the distributional analysis of tax expenditures. In both the second and

third sections, we consider the additional information provided from hypothetical tax

expenditures calculated using TAXSIM. Based on our discussion and analysis, we present

recommendations for reform in the fourth section. The final section concludes.

WHAT’S NORMAL?

A tax expenditure estimate shows the change in Federal income tax revenues due to the

hypothetical elimination of a provision of the Internal Revenue Code. To qualify as a tax

expenditure, the provision must be classified as a divergence from the defined normal income tax

system. As noted above, the Budget Act does not define the reference normal tax. Hence,

10 Among other proposals, the Century Foundation recommended: (1) combined tax expenditure estimates for groups of tax expenditures to account for interaction effects; (2) historical reporting of tax expenditures using comparable methods; and (3) distributional tax expenditure reporting for major provisions. These recommendations mirror some of our own conclusions and suggested reforms, as listed in Section 4 below.

5

building a list of tax expenditures requires analysts to first define what constitutes the normal

income tax. Baseline tax revenues for the estimation exercise can then be generated in reference

to this “normal” tax.11

Defining the normal income tax

Defining the baseline tax structure is inherently a subjective exercise since it assumes

some parts of existing law are “normal,” while others are intentional policy deviations. At a basic

level, defining the normal income tax requires analysts to specify the tax base, the rate structure,

and the tax unit.12 The staffs of the JCT and OTA define the normal structure of the income tax

as one personal exemption for each taxpayer, one exemption for each dependent, the standard

deduction, the prevailing rate structure for the individual income tax, a separate corporate income

tax with one marginal tax rate, and deductions for expenses related to earning income. Consistent

with the tax code, tax brackets differ based on marital status. Thus, the tax unit is the individual

taxpaying unit for the individual tax and the corporation for the corporate tax.

The staff of the JCT defends the inclusion of the personal exemption and the standard

deduction by asserting that these provisions implicitly define a “zero bracket”; the OTA cites tax

administration as a possible reason to include these provisions in the normal tax base (see JCT

2007 and OMB 2008). Neither reason is entirely compelling.13 One could argue that all

provisions that make up the tax liability threshold, or the income level at which a family begins to

pay positive tax, should be part of the normal tax. Under this scenario, the child credit and earned

income tax credit (EITC), for example, would not be tax expenditures because they constitute part

11 The revenue baseline is generated using projections of Federal receipts either from the Congressional Budget Office, for the JCT estimates, or the Office of the Management and Budget, for the Treasury Department estimates. 12 As the paper by Carroll, Joulfaian and Mackie (2008) in this volume makes clear, a provision that is a tax expenditure under one tax base (a comprehensive income tax, for example) may not be classified as an expenditure under an alternative base (a consumption tax, for example). 13 Including the standard deduction for tax administration reasons is appropriate. It is not clear why this argument would apply to the personal exemption.

6

of the zero bracket. Alternatively, one could treat the personal exemption and standard deduction

as tax expenditures, which they would be, for example, if the normal income tax was structured

as the current alternative minimum tax (AMT).

More fundamentally, these uneven distinctions between what is a tax expenditure and

what is part of the normal income tax reveal differences that may be meaningful within a tax law

context but not within an economic context. As in the case of the EITC, a credit is almost by

definition a tax expenditure. However, one could imagine a set of deductions and credits that

generates the equivalent tax liabilities for lower-income taxpayers as the prevailing set of

marginal tax rates. Under existing practice, the set of deductions and credits is a tax expenditure

(for income support) and the set of rates is part of the normal tax system. This discussion serves

to illustrate the subjective nature of tax expenditure estimation and classification.

Present practice is not even entirely consistent with respect to rates of tax. While the

progressive tax structure in the individual tax is considered normal, the graduated corporate rates

that are part of the corporate income tax are treated as a tax expenditure.14 Furthermore, the

reduced rates of tax on long-term capital gains and dividends enacted with the Jobs and Growth

Tax Relief Reconciliation Act of 2003 (JGTRRA) are classified as tax expenditures by the staff

of the JCT but not by the staff of the OTA. As of 2007, these special rates represent the largest

single tax expenditure in the JCT tax expenditure list (JCT 2007).15

The normal income tax versus a comprehensive income tax

14 The staff of the JCT includes the lower rates as a tax expenditure arguing that they provide a tax benefit for small businesses. From a policy perspective, including graduated individual rates in the normal tax structure while at the same time not including graduated corporate rates is consistent policy. The multiple individual rates reflect a desire to tax according to ability to pay. There is no similar ability to pay concept for corporations. 15 OTA has included the reduced rates on dividends and capital gains as part of the normal tax since 2005. The Budget document explains that “In a gradual transition to a more economically neutral tax system under which all income is taxed no more than once, the lower tax rates on dividends and capital gains on corporate equity under current law have not been considered tax preferences since 2005 (OMB 2007, page 299).” As a result, the OTA tax expenditure estimates for the lower rates on capital, step-up in bases, and inside build-up on tax-preferred assets are limited to capital gains from sources other than corporate equity.

7

The Administration budget presentation includes a detailed discussion of how the normal

tax structure differs from a comprehensive income tax (see also Carroll, Joulfaian, and Mackie

2008). We briefly highlight the main differences in this section. As mentioned above, the normal

tax structure used by JCT and OTA includes a separate corporate income tax. Under a

comprehensive income tax, the two systems would presumably be integrated so that all income

would be subject to taxation once at the recipient’s personal tax rate. Although the normal

structure does include a foreign tax credit and dividends received deduction to prevent double

taxation at the corporate level, the present concept does not include any provision to mitigate the

double taxation of income that results from having both corporate and individual tax systems. As

mentioned in the introduction, the staff of the OTA includes an appendix to the tax expenditure

section of the Budget that reports the “double tax on corporate profit” as a negative tax

expenditure. The estimate is negative since the “provision” raises rather than loses revenue

relative to the baseline tax system.

Another major departure from a comprehensive income tax base involves the timing of

taxes. While a comprehensive income tax would tax all income as it is earned, the normal

structure taxes capital gains only upon realization. Thus, deferral of tax on capital gains is not

considered a tax expenditure.16

A comprehensive income tax would levy tax on real as opposed to nominal gains in

capital asset or debt values. But present law fails to provide any indexing of the bases of capital

assets or debt. Both the JCT and OTA follow present law in their normal tax definition.

A comprehensive income base would include accrued and imputed income such as

imputed rent from housing and imputed services from consumer durable goods. Until 2006, the

revenue loss from the exclusion of imputed rent was not reported in either the JCT or OTA tax

expenditure exercises. Starting with the FY 2006 Budget, the Treasury began estimating the tax

16 It is interesting to note that some provisions in the tax code that allow deferral of taxation are identified as such. For example, the deferral of gain on like-kind exchanges and interest on savings bonds are tax expenditures. Concerns of administration and liquidity argue in favor of this inconsistency.

8

expenditure for the exclusion of implicit rental income of homeowners net of depreciation,

interest and taxes. The present treatment creates an inconsistency, however. The OTA estimate

for net imputed rent appears in the regular listing of tax expenditures. But if imputed rent is

treated as a tax expenditure, then the mortgage interest deduction and deduction for taxes on

owner-occupied homes should not be treated as tax expenditures. OTA lists all these provisions

as tax expenditures, which results in double counting. Moreover, the OTA does not report

similar, although smaller, imputed-rent tax expenditures for other durable goods, such as cars.

The individual AMT and passive loss rules create their own complexities with respect to

tax expenditure classification. A comprehensive income tax would (presumably) not include an

AMT and would allow for the full deduction of losses generated from passive activities. The

annual JCT tax expenditure pamphlet explains that the AMT and the passive activity loss rules

are not part of the normal income tax. Instead, they are viewed as provisions that reduce the

magnitude of the tax expenditures to which they apply. Exceptions to the AMT and the passive

loss rules are therefore not classified as tax expenditures by the staff of the JCT because the

effects of the exceptions are already incorporated in the estimates of related tax expenditures.

The OTA lists the AMT and passive loss rules as part of the normal tax system.

Like the individual AMT, the corporate AMT is part of the normal tax structure for the

OTA but not for the JCT. Both agencies consider the prevailing rules providing for carryback

and carryforwards of business net operating losses as a part of the normal tax structure. Under a

comprehensive income tax, however, losses would be fully and immediately deductible. Another

difficult issue for any tax system involves the tax treatment of the recovery of capital costs.

Analysts must specify the “normal” treatment of depreciation for tax purposes. A comprehensive

income tax would provide inflation adjusted allowances for economic depreciation. The OTA

follows this treatment and uses estimates of real, inflation adjusted, economic depreciation in its

normal tax baseline. In contrast, the JCT treats capital cost recovery allowances that are more

9

generous than those provided under section 168(g), which allows straight-line depreciation over a

period that is longer than under the current law accelerated system, as tax expenditures.17

Differences in normal tax baseline across institutions

As should be clear, the definition of the normal tax is not obvious. The normal tax is not

analogous to a comprehensive income tax or the current tax system and changes over time.

Further, JCT and OTA employ slightly different definitions. In addition, as mentioned in the

introduction, the staff of the OTA currently uses two baselines: a normal tax baseline and a

reference tax baseline with the latter being closer to existing tax law.

Table 1 shows the differences between the JCT and OTA definitions of “normal tax” as

well as the difference between the OTA “normal” and “reference” tax baselines. As the table

demonstrates, the JCT methodology uses a somewhat broader definition of the normal income tax

base. Accordingly, the JCT list of tax expenditures includes some provisions that are not

contained in the Treasury list.18 The most striking differences, in terms of the magnitudes of the

estimates, are the omissions from the OTA tax expenditure list of reduced rates on dividends and

capital gains, cash accounting, and the exclusion of the value of Medicare Part A (hospital

insurance), Part B (supplementary medical insurance), and Part D (prescription drug insurance) in

excess of premiums.

The normal tax structure and tax expenditure estimates

Tax policy proposals that change the normal tax structure by definition change the

magnitude of tax expenditure estimates. A decrease in marginal tax rates or an increase in the

17 Prior to 2004, OTA used the JCT methodology. 18 Nonetheless, there are tax expenditures reported by OTA that are not reported by the JCT. Until 2007, the JCT pamphlet has included a list of tax expenditures that are scored by the JCT but not by Treasury (and vice-versa). An interesting difference in particulars but not to the general rule is that Treasury counts a credit against excise taxes that has no effect on income tax liabilities, the alternative fuel and fuel mixture tax credit, as a tax expenditure. JCT provides an estimate for an income tax credit provided for the carrying of excise taxes paid on distilled spirits in wholesale inventories.

10

standard deduction, for example, will decrease the tax expenditure for the mortgage interest

deduction. As a result, both the size and distribution of tax expenditures can be affected by

intentional as well as unintentional policy changes.

Consequences of the standard deduction

Including the standard deduction in the normal tax structure has an important impact on

the tax expenditures for itemized deductions. As explained further in the next section, the

procedure for estimating tax expenditures essentially consists of running taxpayers through a tax

calculator under the baseline normal tax structure with and without the tax provision and

comparing tax revenues. A taxpayer that would not itemize in the absence of a particular

itemized deduction is assumed to claim the standard deduction (this is sometimes called “tax form

behavior”). As a result, the standard deduction reduces the estimate of some tax expenditures

arising from itemized deductions. Some examples of the effect of the standard deduction on the

tax expenditure estimates are shown in Table 2.

The hypothetical taxpayer in Table 2 itemizes and claims a total of $13,100 in itemized

deductions for state and local income taxes, real estate taxes, home mortgage interest, and

charitable contributions. The standard deduction is assumed to be $10,000 and the taxpayer is

assumed to face a marginal tax rate of 20 percent regardless if s/he itemizes. We ignore

complications related to the AMT for now. The first column shows the base case; the next four

columns show the tax expenditure estimates for each of the itemized deductions.

Note first that although the deductions this taxpayer claims for state and local income

taxes and for real estate taxes are different, the tax expenditure estimates are identical. This is

shown in columns (2) and (3). Both the itemized deduction for state and local income taxes and

for real estate taxes are large enough that if either were removed, the taxpayer would claim the

standard deduction. The increase in taxable income is simply the difference between the

11

taxpayer’s itemized deductions ($13,000) and the standard deduction ($10,000). Thus, the

revenue gain from eliminating either deduction is 20 percent of $3,100 or $620.

Now consider what happens if the mortgage interest deduction is disallowed. In this case

the tax expenditure estimate is based on the full amount of this taxpayer’s mortgage interest

payments and the standard deduction has no effect on the estimate (see column 4). Our example

also shows how tax expenditure estimates vary with the magnitude of the standard deduction. For

example, increasing the standard deduction by $2,000 makes all of this taxpayer’s itemized

deductions “marginal” in the sense that the taxpayer takes the standard deduction in the absence

of any one of their itemized deductions. As a result, the tax expenditure is the same for all four

deductions.

The last row of table 2 shows clearly that the tax expenditure estimate of an itemized

deduction does not reflect how revenues would change if the itemized deduction were an “above

the line” deduction or a statutory adjustment. Of course, this is not the exercise that tax

expenditure estimates were meant to address but is one that can be of interest to policy analysts.

Our example ignored the phase-out of itemized deductions under current law which is

part of the normal tax baseline (and has been since the so-called Pease provision was put in place

in 1990). The phase-out reduces expenditure estimates of itemized deductions for taxpayers over

certain income thresholds under the regular tax but not under the AMT leading to complicated

interactions between the standard deduction, itemized deductions, and the AMT. Barthold et. al.

(1998) show how these phase-outs, phase-ins, and other income tests with respect to certain tax

provisions can create conditions under which a taxpayer’s effective tax rate can exceed their

statutory marginal tax rate, with additional complications for revenue and tax expenditure

estimation.

Consequences of the AMT

12

The AMT reduces the value of tax preferences by “recapturing” them. The AMT also

affects the value of certain tax preferences by changing effective marginal tax rates.19 To

demonstrate the importance of the AMT, we use TAXSIM to calculate how a set of tax

expenditures estimates vary across different AMT policy scenarios. TAXSIM is the individual

income tax simulation model operated by the NBER.20 Like the models used by JCT, OTA, and

some think tanks (the Tax Policy Center, notably), TAXSIM is essentially a large Turbo-Tax-

style calculator aggregating tax form results over thousands of simulated, statistically-weighted

taxpayers.21 The underlying data for TAXSIM is the IRS Statistics of Income (SOI) public-use

file. JCT and OTA have access to the set of confidential IRS taxpayer data, which is both larger

in terms of sample size and more detailed with respect to tax variables reported.

We calculate tax expenditures for the years 2000-2011 under three AMT scenarios: (i)

current law, (ii) current law with no AMT, and (iii) current law in the absence of a patch for the

AMT. By “current law” we mean the law in place in the year of the estimate so that the 2004 tax

expenditure for the child credit, for example, uses the Internal Revenue Code in place (and thus

the normal tax structure) for 2004. The “patch” refers to recent annual adjustments to the AMT

exemption amounts for inflation to prevent a large increase in the number of taxpayers who

would otherwise pay AMT. By “no AMT patch” we mean that the patch applied in 2001 to the

AMT is no longer indexed.22

Figure 1 shows the tax expenditure for the mortgage interest deduction, the state and

local income tax deduction, and the property tax deductions available to itemizers from 2000 to

2007 under our three AMT scenarios. We focus on the period through 2007 because from 2008 19 For example, mortgage interest paid, with the exception of home equity loan interest not used for home improvement, may be deducted under the regular income tax and the AMT. Moving from the income tax paying-status to AMT-paying status implies changing the applicable marginal tax rate and thus the value of the tax expenditure. 20 We thank Daniel Feenberg for his exceptional efforts in employing TAXSIM for this project. 21 As discussed further in the next section, tax expenditure estimates do not allow for any taxpayer behavior. Revenue estimates produced by JCT and OTA allow for microeconomic behavior while holding the size of the economy constant. As a result, information beyond what is produced by a tax calculator is required for revenue estimates. 22 The TAXSIM estimates are available upon request from the authors.

13

onward the “current law” and “current law with no AMT patch” scenarios are the same. The

decline in the tax expenditure estimates for all three deductions between 2001 and 2003 is a result

of the decrease in rates enacted with the Economic Growth and Tax Relief Reconciliation Act of

2001 (EGTRRA).

Note that the tax expenditure for the property tax and state and local income tax

deduction would have remained near the 2003 level if the AMT exemption levels had not been

“patched.” This is because these itemized deductions are preferences under the AMT. Figure 1

shows that the AMT has had a substantial effect on these tax expenditures. With no AMT, both

tax expenditures would have increased substantially between 2003 and 2007: from $43.4 billion

to $60 billion for the state and local income tax deduction and from $23.8 billion to $32.7 billion

for the property tax deduction.

The effect of the AMT on the mortgage interest deduction is much different. Because

most mortgage interest expenses may be deducted under the AMT, the TAXSIM estimates for the

“current law” and “current law with no AMT” scenario are almost the same. Whether the AMT

is patched, however, has a substantial impact on the expenditure estimate. It turns out that the tax

expenditure for the mortgage interest deduction is higher when the AMT is not patched than

under current law with the patch (see also Burman et al). With no patch, more taxpayers end up

on the AMT. Because there is no standard deduction under the AMT, there is no carve-back to

decrease the tax expenditure of the provision. In addition, under the AMT there is no phase-out

for itemized deductions, which again increases the expenditure estimate. Finally, the rate at

which mortgage interest (and other itemized deductions) are deducted under the AMT and under

the regular tax differ. Using TAXSIM, Feenberg and Poterba (2004) calculate that the weighted

average marginal tax rate for the mortgage interest deduction is about two percentage points

greater under the AMT than without the AMT in 2007, for example.

The AMT affects a wide range of tax expenditures. Figure 2 shows the impact of the

AMT on the itemized deductions examined above for 2007 as well as on three additional tax

14

provisions: the reduced rates for dividends and capital gains, the child tax credit and the Lifetime

and Hope education credits. We chose these provisions for illustrative purposes.

The reduced rates on dividends and capital gains apply under the AMT, but due to an

interaction with the AMT exemption amount, the benefits of the rates can be reduced (see

Leiserson 2007 for a detailed example). When a taxpayer realizes capital gains, s/he increases

AMT taxable income which in turn decreases the AMT exemption through the exemption phase-

out. The phase-out raises the effective rate on capital gains and dividends.

According to our TAXSIM calculations, the 2007 tax expenditure for the reduced rates

on dividends and capital gains is $88.9 billion. With no AMT to carve back the provision, the tax

expenditure increases to $92 billion; a not inconsiderable difference. With no patch of the AMT

exemption amounts, the tax expenditure is slightly higher than the current law estimate and lies

between these two estimates at $89.1 billion because of the reduced ability of the AMT to

recapture the benefits of the reduced rates.

The treatment of tax credits under the AMT has changed recently. For taxable years

beginning before 2007, nonrefundable individual credits were allowed against both the regular

tax and AMT. Starting in 2007, these credits are effectively disallowed --- they may only be

claimed to the extent that the individual’s regular income tax liability exceeds the individual’s

minimum tax liability (see JCT 2007 for details). The adoption credit, child credit, and saver’s

credit, however, are exceptions to this rule and may be claimed against both the regular tax

(reduced by other nonrefundable personal credits) and the AMT.

Figure 2 shows clearly that the AMT has an important and uneven impact on tax

expenditures. Much has been written concerning the uncertainty created by the AMT (see

Ackerman and Altshuler 2006, for example). This uncertainty spills over to tax expenditure

estimates that are presented only for present tax law baselines and not alternative scenarios that

may be more realistic such as an extension of the AMT patch. An open question that we address

later is whether tax expenditures should be calculated under a variety of AMT scenarios.

15

Hypothetical tax expenditures

We have noted that what is considered “normal” is subject to judgment. In this section,

for illustrative purposes, we consider some hypothetical tax expenditures that would arise under

different definitions of the normal tax. These “new” expenditure estimates provide information

regarding the cost of the structure of the prevailing tax system.

Suppose first that the standard deduction and personal exemptions were classified as tax

expenditures (listed as income support policies, for example). Using the TAXSIM model, we

find that the personal exemption estimate is $129 billion in 2007 and represents 12.8 percent of

baseline revenues. Clearly, this deviation from a pure income tax with no deductions or

exclusions is large relative to other policy divergences in the tax code.

The standard deduction is more difficult to estimate using TAXSIM since it requires

imputations for itemized deductions for taxpayers that do not itemize under present law. If we

assume that in the absence of the standard deduction there exists a floor on the itemized

deductions equal to the current standard deduction, then the hypothetical tax expenditure for the

standard deduction is $75 billion for 2007, a value that clearly rivals other major existing tax

expenditures.

Similarly, the progressive structure is a set of preferred tax rates, and special rates are

considered a tax expenditure under the Budget Act. Therefore, an argument can be made that this

system is itself a tax expenditure. In fact, as noted earlier, one could construct a set of

deductions/credits under a pure flat tax that generates the same tax liabilities as the progressive

rates create under present law. The set of deductions/credits would be tax expenditures; the set of

progressive rates would not. Nonetheless, most tax analysts would agree that the ability-to-pay

principle, and thus the progressive system of rates, is an important element of the existing income

tax. However, if one were to report a tax expenditure for this component, it would be similar to

the following exercise.

16

Assume that for the purposes of this hypothetical tax expenditure, the maximum statutory

marginal income tax rate in a given year (35 percent for our purposes) is the “normal” rate. All

other rates under present law are therefore preferential rates, assigned on the basis of income type

(e.g. wages or capital gains) or other criteria (adjusted gross income of the taxpayer). These

preferred rates are tax expenditures.

We estimate tax expenditures for each of the rates using 2005 SOI data. 23 Figure 3

shows our expenditure estimates for the 2005 individual tax brackets. The tax expenditure for the

10 percent bracket (the estimate for taxing income in this class at 10 percent rather than 35

percent) is equal to $244.7 billion, a number that substantially exceeds most existing tax

expenditures. The estimate for the 15 percent rate is $344 billion, for the 25 percent rate is $72.1

billion, and for the 28 percent rate is $17 billion. In total, the hypothetical tax expenditure for

“income support for the progressive system of rates” is $827.4 billion per year, an estimate that

exceeds almost all official tax expenditures.

Combined with the estimates for the standard deduction and the personal exemptions,

these calculations provide information regarding the amounts that are dedicated in tax

expenditure terms to the progressive system of rates and reveal the relative size of this important

policy element to other, existing tax expenditures. As with other tax expenditures, the

distributional aspects of this hypothetical tax expenditure are revealing. Appendix Figure 1

shows that the benefit of the progressive system of rates falls roughly on middle-income

taxpayers.

Comparing estimates over time

23 One could imagine other methods of calculating these tax expenditures. Instead of using the maximum statutory rate as the baseline rate, one could use the average marginal rate or the average effective tax rate. However, doing so would require reporting negative tax expenditures for income taxed at above average rates, which presents its own conceptual issues that are discussed in more detail below. Alternatively, one could use the lowest rate as the base, and calculate negative tax expenditures, or use the revenue neutral rate.

17

Many researchers have studied how tax expenditures have changed over time.24 But, as

noted above, the baseline “normal” tax changes when tax policy changes alter any provisions of

the prevailing tax structure. Hence, Congress implicitly changes policy towards tax-favored

activities when it changes tax rates, the standard deduction, the AMT, or any other part of the

normal tax structure. To demonstrate how policy changes affect tax expenditure estimates, we

use TAXSIM to calculate a set of tax expenditures under present law in place in each year and

under pre-2001 law (we call this “constant” law since it holds constant the provisions in place

prior to EGTRRA). In our discussion, we focus on the mortgage interest, state and local income

tax, and property tax deductions.

Figure 4 shows the effect of tax policy changes enacted from 2001 onward on the growth

of the tax expenditure estimate for the mortgage interest deduction. Our calculations indicate that

this tax preference to housing increased by 14 percent between 2000 and 2007. In the absence of

the 2001 and 2003 tax policy changes (under “constant law”), the tax expenditure estimate would

have increased by 26 percent. Note that between 2008 and 2010, the tax expenditure estimates

are quite similar. This is due to the AMT which is not patched under current law for those years,

and, as we have seen, actually increases the tax expenditure for this itemized deduction.

The evolution of the tax expenditure for state and local income taxes under constant (pre-

2001) law and current law is shown in Figure 5 and is striking. Between 2000 and 2007, this

expenditure decreased by about 1 percent under current law but would have increased by 36

percent under constant law. The noticeable decrease from 2007 to 2008 is due to recapture

effects from the AMT, which is not patched for 2008 under our present law baseline. The

difference in the benefits for the property tax deduction under actual law and the constant law

baseline, shown in Figure 6, is also dramatic. The tax expenditure increases by 24 percent

between 2000 and 2007. Under constant law this expenditure would have grown by 72 percent.

This exercise illustrates clearly how tax policy changes can effect the growth of tax expenditures. 24 See, for example, US GAO 2005, Neubig and Joulfaian 1988, Toder 1998, and Steuerle 2004.

18

Sunsets, budget windows and timing issues

Many major and minor elements of the tax code are scheduled to sunset in the near

future. In fact, most of the major 2001 and 2003 tax cuts are scheduled to expire on January 1,

2011, a date some observers have called “tax policy doomsday.” The scheduled sunsets lead to

some interesting patterns of tax expenditure estimates as we have already seen in Figures 4, 5,

and 6. Reading a tax expenditure table across time requires fairly sophisticated knowledge of

U.S. tax policy and is a bit like reading tea leaves. Consider the tax expenditure estimate for the

state and local income tax deduction under current law. The estimates for 2000-2011 assume that

the AMT is not patched for 2008-2011 and the 2001 and 2003 tax cuts expire as scheduled.25 It

is hard to think of this pattern of tax benefits as being equivalent to an expenditure program

outlay given how it fluctuates with tax rules.26

Sunsets can also give rise to negative tax expenditures. For example, the 2007 JCT tax

expenditure publication reports negative tax expenditure estimates for the tax code section 198

brownfield expensing provision for individual taxpayers (i.e. pass-through entities and small

businesses) for fiscal years 2009 through 2011. This is due to the fact that the provision expires

under present law at the end of 2008. Expensing accelerates qualified brownfield remediation

cost deductions. The acceleration generates a positive tax expenditure in the year of expensing

with a corresponding set of negative tax expenditures due to the loss of deductions in future

years. An expensing provision with no sunset would not typically reveal these negative tax

expenditures because the expensing behavior in each fiscal year would exceed the loss of the

25 In particular, the marginal income tax rate cuts, including the creation of the 10% bracket and the reduction of the maximum rate from 39.6% to 35%. 26 Nonetheless, expenditure programs have their own quirks with respect to reporting, including the annual appropriation process and some baseline issues.

19

otherwise permitted expense deductions.27 Negative tax expenditures that arise in this manner

appear for other provisions, including section 179 small business expensing, GO-Zone bonus

depreciation, and Indian reservation accelerated depreciation.

Another class of examples of negative values appearing on tax expenditure tables is the

set of expiring business tax credits, such as the credit for fuel cell and microturbine power plant

installation in OTA’s 2008 tax expenditure listing. Here, the provision provides a credit to the

qualifying taxpayer, but the credit also requires a corresponding reduction in basis of the asset.

This consequently reduces future depreciation deductions and raises gains taxation, if applicable,

in future years. Beyond the sunset of the credit, this gives rise to larger tax receipts than in the

baseline and therefore a negative tax expenditure.

Nonetheless, as a general rule, negative tax expenditures are not presently reported as a

matter of convention.28 In the examples noted above, these negative estimates arise only within a

few or even single years.29 Over the full reporting period for the tax expenditure, in general no

negative tax expenditures are currently estimated and reported in tax expenditure budgets

produced by either the JCT or OTA.30

This raises a fundamental question: what is a negative tax expenditure? For many, a

negative tax expenditure represents a disproportional statutory tax burden on a particular kind of

income or economic activity. As an outlay equivalent, it may be thought of as a fee or transfer

payment from the taxpayer to the Treasury. However, for others, a negative tax expenditure is

simply a tax and thus no special accounting is necessary. Indeed, the Budget Act does not refer to

27 However, the original rules for the brownfield expensing provision was limited to certain locations. If all such locations were redeveloped, then even a permanent version of the original rules would have exhibited negative tax expenditures for individual years. We thank Tom Barthold for noting this point to us. 28 JCT (2008) announced that they will report negative tax expenditures in the future. 29 However, this raises the question of whether a negative tax expenditure should continue to be reported for an expired provision that has only a negative tax expenditure for the budget window. 30 One exception illustrates the previous explanation of temporary provisions giving rise to negative tax expenditures. The 2007 OTA report provides a negative tax expenditure over the reporting period (2008-2012) for the deferral of gain allocable to disposition of electrical transmission property due to FERC-mandated restructuring.

20

negative tax expenditures. Nonetheless, a certain respect for symmetry requires that more

thought and analysis be given to negative tax expenditure reporting.

OTA lists several examples of tax rules that may be considered negative tax expenditures

under a comprehensive income tax. These include the passive loss rules, the restrictions on

carrybacks and carryforwards for net operating losses, and restrictions on the deductibility of

capital losses. For example, owners of a personal residence are not permitted to deduct a loss

from the sale of the property. The gain exclusion for principal residences is reported as a positive

tax expenditure, but the corresponding restriction on loss deductions is not. There are other

theoretical negative tax expenditures that are worth considering. For example, the double

taxation of corporate income may be considered a tax expenditure. And, in fact, beginning in

2004, the OTA has been reporting the double taxation of corporate profits as a negative tax

expenditure in an appendix to the tax expenditure presentation in the budget. As another

example, the taxation of gains due to inflation might be considered a negative tax expenditure.

Alternative baselines

As this section has made clear, different “normal” tax structures will generate different

tax expenditure estimates. The staff of the OTA has studied how the expenditure budget would

change under a comprehensive income or consumption tax base using the current rate structure

(see OMB 2008 and Carroll, Joulfaian and Mackie 2008). It is also interesting to consider the

impact of the current progressive rate structure on expenditure estimates. One way to explore this

issue is to use a flat tax structure to estimate tax expenditures. We ran some experiments with

TAXSIM replacing the normal marginal tax rates (and special rates on dividends and capital

gains) with a flat rate of 19 percent, the rate that most closely corresponds to a revenue neutral

rate according to TAXSIM for 2008.

Again, for expository purposes, we focus on the tax expenditures for the mortgage

interest, state and local income tax, and property tax deductions. In addition to current law and

21

pre-EGTRRA law estimates, Figures 4, 5, and 6 include series showing estimates for our flat rate

experiment. The growth of the mortgage interest deduction over the period, shown in Figure 4, is

similar under current law and current law with the flat rate because of its relative lack of

interaction with the AMT. The estimates plotted in Figures 5 and 6 reveal the more significant

interactions between the AMT and the state and local income tax and real property tax

deductions. The flat rate experiment shows a relatively smooth pattern over the time period while

the current law experiment illustrates the significant carve-back of these deductions by the AMT.

Also interesting is the fact that the flat rate estimate exceeds the 2010 and 2011 estimates for pre-

EGTRRA law due to the relatively lower marginal rates faced by taxpayers deducting property

taxes.

Summary

It is clear from the preceding discussion that tax expenditure reporting involves numerous

definitional and interactive complications that distinguish it from outlay or government

expenditure accounting for which tax expenditures are intended to be comparable. Debates

regarding the appropriate “normal” income tax structure, timing issues, interactions with the

AMT and standard deduction, as well as more fundamental issues regarding the correct income

definition suggest that tax expenditure reporting must be examined within the context of certain

standing and generally accepted assumptions regarding tax policy. Nonetheless, as the following

section illustrates, even the more mechanical estimation process, done under a given set of such

assumptions, is complicated by common misunderstandings regarding the methodology used to

generate estimates of tax expenditures.

SCORING TAX EXPENDITURES

Estimating Considerations

22

The economists of the JCT and the OTA employ certain standing assumptions when they

calculate tax expenditure estimates. First and most importantly, unlike official revenue estimates,

tax expenditure estimates do not incorporate any change in taxpayers’ economic behavior.

Revenue estimates include forecasts of behavior that will change, including consumption,

investment and other economic actions. These modeled micro-dynamic responses allow

taxpayers to respond to changes in after-tax prices and other tax-related incentives. However, as

is well known, revenue estimation does not allow macroeconomic feedback or changes in

national income in response to changes in tax policy.

In contrast, tax expenditure estimates include neither the micro-dynamic response nor

macroeconomic feedback. As discussed in the previous section, a tax expenditure is simply the

change to the existing baseline or forecast of a particular line item of tax revenue. Therefore, it is

important to note, as some analysts fail to do, that a tax expenditure estimate is not a revenue

estimate.

For example, the tax expenditure for the mortgage interest deduction is calculated by

summing for all itemizing taxpayers the amount of mortgage interest paid times the applicable

marginal income tax rate applied against itemized deduction amounts. As noted earlier, there is

an interaction with the standard deduction due to the itemization decision. However, the tax

expenditure estimate does not allow the taxpayer to modify the own-rent decision regarding

housing, nor does it allow the taxpayer to change debt and investment allocation decisions. It is

reasonable to assume that if Congress were to eliminate the mortgage interest deduction, many

taxpayers would reduce their holdings of low-yield assets to reduce mortgage debt holdings due

to the change in after-tax mortgage interest rates.31 While a revenue estimate would reflect these

actions, the tax expenditure estimate does not. For most cases, the tax expenditure estimate tends

to be larger than the corresponding revenue estimate because of this assumption.

31 Poterba and Sinai (2008) examine this tax expenditure in detail.

23

As another example, the tax expenditure estimate for the Hope credit for student expenses

does not allow taxpayers to change education decisions in order for the taxpayer to qualify for

other tax incentives. There are numerous education incentives in the Internal Revenue Code,

some with overlapping qualifying criteria. The interaction of these incentives is discussed more

below.

To examine the difference between a revenue estimate and a tax expenditure estimate,

consider the following partial equilibrium example for an economic action, Q, with price, P.

Suppose that the marginal cost curve (MC) includes a broad-based tax, t. The marginal cost and

marginal benefit cost curve yield the after-tax partial equilibrium. Further suppose that there

exists a separate tax provision that provides a tax expenditure targeted directly to this market.

The tax expenditure takes the form of a per-unit subsidy in the amount of p2 minus p1, thereby

shifting the applicable marginal cost curve to MC`. The tax expenditure increases the amount of

the activity undertaken from q* to q. Under the prevailing tax rules, the tax expenditure induced

quantity, q, is the baseline quantity for tax expenditure analysis purposes.

24

The tax expenditure estimate of the incentive is equal to (p2-p1)q.32 However, the

revenue estimate of repealing this tax expenditure is equal to (p2-p1)q – t(q-q*). The revenue

estimate is lower because it includes a revenue loss associated with the microdynamic response: a

reduction in the amount of the activity undertaken by the taxpayer from q to q*.

It is difficult to provide an example of a set of official estimates illustrating the difference

between a revenue and tax expenditure estimates, typically because tax expenditure estimates are

calculated for large provisions or bundles of small, related provisions of the tax code. Revenue

estimates typically are issued for smaller, more detailed policy changes associated with these

provisions. For example, in 2005 the JCT published a revenue estimate for repealing the

deduction for home equity loan interest paid (JCT 2005). With respect to this provision, only the

total tax expenditure for the mortgage interest deduction is publicly available. In contrast, the

repeal would have affected only a portion of the mortgage interest deduction, so a direct

comparison between the tax expenditure and the revenue estimate cannot be made. Despite these

data limitations, it is certainly the case that the 2005 revenue estimate included a number of

behavioral responses to the proposed policy change and was therefore lower than its hypothetical

tax expenditure.

An example that can be used to illustrate this point is the tax exclusion for employer-paid

life insurance. The JCT reports that the tax expenditure for this provision in years 2009, 2010

and 2011 is approximately equal to $2.7 billion. However, in the 2007 CBO Budget Options

report, the JCT reported that the revenue estimate for this provision is only equal to $2.1 billion in

2009, $2.2 billion in 2010, and $2.3 billion in 2011. The smaller revenue estimate is consistent

with what theory would predict. The tax incentive for employer-provided life insurance increases

its use, and repealing that tax expenditure would not result in a full capture of the benefit because

32 We assume here that the tax expenditure does not take the form of an itemized deduction to abstract from any interaction with the standard deduction.

25

of shifting of taxpayer behavior. Therefore the revenue estimate should be lower than the tax

expenditure.

As noted earlier, there is one notable exception to the general rule that tax expenditure

estimates do not incorporate taxpayer behavior: tax form behavior. In this sense, tax expenditure

estimates automatically include a realistic accounting of most tax form interactions, including

whether to itemize or not. This tax form behavior creates the possibility of changing the relative

sizes of various tax expenditures, as our earlier example from Table 2 illustrated. In general, tax

expenditures of large provisions generate relatively higher estimates than tax expenditures of

relatively smaller provisions because of the itemization decision. Given that tax expenditures are

often compared to one another, this biasing of tax expenditures in terms of size may be

problematic for tax policy analysts. On the other hand, the recapture produced by the standard

deduction is clearly an important component of the revenue estimate and thus this estimate may

in fact be more realistic.

However, there are tax incentives that have overlapping qualifying criteria for which only

one tax incentive may be claimed as a matter of law. How should such provisions be estimated

for tax expenditure analysis? Because the qualifying criteria are similar, the taxpayer may qualify

for another tax expenditure with no change in economic behavior, if the one claimed under

current law were to be eliminated. However, this type of tax form behavior is not likely to be

automatically captured by an Individual Tax Model (ITM), and thus can lead to differing

estimates.

As an example of the differences in scoring conventions for this situation, consider the

JCT and OTA tax expenditure estimates for the tax credits for post-secondary education (HOPE

credit and the lifetime learning credit).33 The credits have differing but similar qualifying criteria,

such that if one credit were repealed, then some taxpayers could claim the other credit. The

33 Another interesting example of overlapping provisions involves the treatment of export income. Prior to the Jobs Act, U.S. multinational corporations received favorable treatment of income from sales abroad through the export source rules and the foreign sales corporation provisions.

26

differences in the JCT and OTA estimates for these provisions demonstrate that they indeed use

different conventions for this estimate. The fiscal year 2008 estimate from JCT for both credits is

$4.4 billion. The OTA estimate is broken out for each credit, with $3.4 billion for the HOPE

credit and $2.2 billion for the lifetime learning credit. It is clear that the JCT economists are

allowing taxpayers to substitute between the two credits, when the taxpayers are otherwise

qualified and one credit is hypothetically repealed for tax expenditure estimation purposes. This

produces some recapture, through increased use of another tax incentive, and a smaller tax

expenditure estimate. In contrast, the OTA estimates examine the credits in isolation and do not

include such tax form behavior. Nonetheless, for both estimates, it is important to note that this

tax form behavior does not equate to change in economic behavior. There is no increase or

decrease in the underlying qualifying activity of higher education expenditures by the taxpayer.34

There are many other overlapping provisions in the tax code. For example, the major

savings incentives, traditional and Roth IRAs and 401(k)-style accounts, reward a similar

underlying economic behavior: saving. The report of the President’s 2005 Advisory Panel for

Federal Tax Reform notes that there are at least a dozen tax-preferred options in the current code.

Should the tax expenditure estimates for such provisions include interactions effects? Or should

they be reported in isolation in order to more closely adhere to the appropriations equivalent

concept? We take this issue up again in the next section.

The interaction of tax expenditures raises a known issue that is routinely ignored with

respect to the reporting of bundles of tax expenditures. Tax expenditures cannot be summed.

Because of the previously identified issues concerning itemization and other tax form behavior,

summing of tax expenditure estimates often results in double counting and biased estimates. For

example, Hungerford (2006) performed a simulation in which twelve selected tax expenditures

were eliminated in isolation and then simultaneously. Hungerford finds that the sum of the

34 There may be other interactions in this example with the tuition and fees deduction that may be claimed on the front of the 1040 individual tax form.

27

individual tax expenditure estimates was 17.5 percent higher than the tax expenditure calculation

for simultaneous repeal of the twelve provisions. The Government Accountability Office asked

OTA to conduct a similar exercise with five major itemized deductions (U.S. General

Accountability Office 2005).35 The analysts at OTA found a 25 percent difference between the

simultaneous estimate and the sum of the individual tax expenditures. Despite this, researchers

have produced reviews of tax expenditures that rely on summing tax expenditures to present an

aggregate picture of the role tax expenditures play in the federal government’s budget.36

Using TAXSIM, we calculated the tax expenditure for two bundles of tax expenditures.

For 2007, TAXSIM reports a tax expenditure for the mortgage interest deduction of $82.4 billion

and $25.8 billion for property taxes for owner-occupied homes. Summing these two tax

expenditures yields $108.2 billion. However, the tax expenditure estimated simultaneously for

these two provisions generates an estimate of $99.1 billion, a decline of 8.4 percent. As a second

and somewhat different example, the 2007 tax expenditure for the state and local income tax

deduction is $47.4 billion. Summing with the tax expenditure for the property tax deduction

yields $73.2 billion for 2008. The simultaneous estimate of these tax expenditures produces

$74.8 billion, representing an increase of 2.1 percent. This unusual case, where the simultaneous

repeal of two tax expenditures yields a larger estimate than the sum of their individual tax

expenditures is due to the AMT. As noted earlier, the AMT can also produce interaction effects,

particularly given the combined effect of lower tax rates and unindexed AMT exemption

amounts, which is forecasted to increase the number of AMT-payers for future tax years. In the

absence of the AMT, the 2007 tax expenditure for the state and local income tax deduction is $60

billion and the tax expenditure for the property tax deduction is $32.7 billion. The simultaneous

35 The five itemized deductions include charitable contributions, home mortgage interest expenses, state and local income taxes, state and local property taxes, and medical expenses. 36 See, for example, Neubig and Joulfaian (1988), Toder 1998, Steuerle (2004), and GAO (2005), and most recently Weiner (2008).

28

estimate of these tax expenditures is $95.5 billion, representing a decrease of 3.7 percent from the

sum of the two tax expenditures ($92.8 billion).

Putting aside the complexities caused by the AMT, these examples suggest that the

interaction effects among tax expenditures, particularly for large tax provisions, can be

significant. Given these estimates, there would be certain benefits for JCT and OTA reporting of

bundles of policy-related or issue-related tax expenditures. For example, an estimate could be

reported for all housing-related tax expenditure or other similarly-related bundles of tax

incentives.

Alternatively, as estimated by Burman, Geissler and Toder (2008), bundles of tax

expenditures could be grouped by their tax-form character (exclusions, itemized deductions,

refundable credits, special rates, etc.). For 2007 estimates, Burman et. al. find that the tax

expenditure for the sum of income exclusions (life insurance contributions, retirement benefits,

and other exclusions) is six percent higher than the sum of the individual tax expenditures. They

attribute this effect to the progressive character of the income tax rates. When tax exclusions are

consider simultaneously (or “stacked” in the ITM jargon), the effective marginal income tax rate

of the taxpayer increases due to greater amounts of taxable income or AMT-paying status. For a

bundle of itemized deductions (mortgage interest, state and local taxes, charitable contributions,

medical expenses and casualty losses), they find the opposite effect. The tax expenditure of the

set is 15 percent smaller than the sum of the individual tax expenditures because of the recapture

effect produced by the standard deduction.

We consider an experiment that bundles the top tax expenditures. The idea is to use a

variant of tax expenditure estimation (a bundled estimate) to gauge the revenue gains from base

broadening. We are able to estimate thirteen major expenditures with TAXSIM including the

mortgage interest deduction ($82.4 billion), the deduction for unreimbursed medical and dental

expenses ($8.3 billion), the deduction for state and local income taxes ($47.4 billion), the

deduction for property taxes on owner-occupied homes ($25.8 billion), the deduction for

29

charitable contributions ($42.7 billion), the earned income tax credit ($41.4 billion), the

refundable child tax credit ($46.2 billion), the lifetime earning and Hope credits ($6.6 billion), the

child and dependent care credit ($2.6 billion), the tax-exemption on municipal bonds ($17.6

billion), the preferred tax rates on capital gains and dividends ($88.9 billion), the tax preferences

for IRAs and Keogh plans ($4.0 billion), and the exclusion for certain pensions and annuities

($17.6 billion).37 Summing these tax expenditure yields $431.4 billion. Simultaneously

estimating the repeal of these provisions produces a 2007 estimate of $387.7 billion, a difference

of 10.1 percent.

It is worth noting the size of this estimate. In 2007, TAXSIM forecasts a total of $1.02

trillion in individual income tax receipts. Hence, these 13 items represent about 38 percent of this

total. As these are tax expenditure estimates, without actual revenue estimates it cannot be

calculated the extent to which income tax rates could be reduced if these provisions were

eliminated. Nonetheless, these expenditures represent a considerable amount of tax base

narrowing. And without the AMT to carve back the benefits of many of these expenditures, the

TAXSIM forecast for the 13 items increases to $453.4 billion or 46 percent of tax revenues. It is

also interesting to compare how the size of the revenue loss due to this group of expenditures has

evolved over time. Under our current law scenario, the share of revenues accounted for by the

top 13 individual tax expenditures has increased from about 33 percent in 2000 to 38 percent in

2007. Using our constant law scenario, or pre-EGTRRA rules, the percentage increases by only

one percentage point to 34 percent in 2007. With no AMT, the percentage goes from about 35

percent in 2000 to 41 percent in 2007.

Tax Expenditures Distributional Considerations

37 Pensions and annuities not included in AGI includes only has social security benefits due to data limitations on the SOI Public Use File.

30

As with any appropriation item of the federal budget, the distributional aspects of tax

expenditures are important, particularly for tax provisions attempting to achieve a policy

objective beyond raising revenue for the government. The best source of information regarding

the distribution of government tax expenditure estimates is the JCT tax expenditure publication.38

In each edition, the JCT presents the distribution of major tax expenditures: mortgage interest

deduction, student loan interest, education credits, child tax credit, untaxed social security and

railroad retirement benefits, child care credit, earned income credit, charitable contribution

deduction, state and local tax deduction (income, sales and personal property), medical expense

deduction and the owner-occupied home property tax deduction.39

As with other forms of distributional analysis, it is important to select the appropriate

income classifier. JCT uses an expanded income classifier which is equal to adjusted gross

income (AGI) plus the following items: tax-exempt interest, employer contributions for health

and life insurance, employer share of payroll taxes, workers’ compensation, nontaxable Social

Security benefits, the insurance value of Medicare benefits, alternative minimum tax preference

items, and excluded income of U.S. citizens living abroad.

Beyond the income classifier concept itself, it is also important to note that the JCT

distributional methodology uses the unit of analysis provided by the SOI data: the tax unit. With

38 Congressional Budget Office reports are another source of distributional information for tax expenditures. The most recent example is a February 2008 report on the deductibility of state and local taxes (CBO 2008). Outside of the government, the Tax Policy Center (TPC) uses their ITM to distribute a variety of individual tax expenditures including, among others, the child tax credit, earned income tax credit, tax benefits for health insurance, and special rates for capital gains and dividends. While the TPC methodology follows the JCT approach, there are some differences. The TPC use a measure of cash income that differs somewhat from the JCT expanded income detailed above. Cash income is AGI minus taxable state and local tax refunds, plus total deductions from AGI, non-taxable pension income, tax-exempt interest, non-taxable social security benefits, cash transfers, worker’s compensation, employer’s contribution to tax deferred retirement savings plans, employer’s share of payroll taxes and corporate tax liability. The inclusion of the employer share of payroll taxes and corporate taxes puts income on a pretax basis. Another difference from JCT is that the TPC examines the impact of tax expenditures on after-tax income, as well as the share of the tax benefit received and the size of the tax benefit. 39 In addition to being relatively large, these provisions have the benefit that the data requirements necessary to confidently estimate the distributional aspects of the tax expenditures are available in the confidential SOI data. For other large tax expenditures, the data and off-ITM model estimations provide less reliable methods of determining distributions.

31

the possibility of multiple tax returns being filed by households or families, as these concepts are

defined in Census data for example, there is some difficulty in using this analysis to make

comparisons to other kinds of distributional analysis. This is particularly true for married

taxpayers filing separate returns. Furthermore, there can be a difference between the statutory

incidence of a tax expenditure and the economic incidence of the tax provision. For example, the

mortgage interest deduction is claimed on an individual tax form, but the underlying deduction

represents expense allocable to investment for a household.

We use the SOI Public Use File for 2004 and TAXSIM to illustrate simple distributional

analysis of the state and local income tax deduction, some grouped expenditures, and a major

expenditure that is not distributed by JCT, the special rates on dividends and capital gains. As

before, we consider three scenarios: current law, current law with no AMT and current law with

no AMT patch. We use a modified version of cash income that can be calculated using

information from TAXSIM as an income classifier.40 To simplify the presentation, we show

results only for 2004 and by income decile (instead of also by income class).41 One goal of this

exercise is to demonstrate how incremental changes in presentation can increase available

information regarding tax expenditure. To this end, we follow the JCT approach and calculate

the distribution of benefits and not some other possible distributional measures such as the effect

of the expenditure(s) on after-tax income.

We begin by showing the distribution of tax liability in Table 7. As is well-known, tax

liability is concentrated in the top decile. Somewhat surprisingly, the distribution of tax liability

changes little under the two alternative AMT scenarios. Table 8 shows the distribution of the tax

40 Cash income is AGI minus state and local tax refunds, plus tax-exempt interest, non-taxable Social Security benefits, deductions for IRA contributions, student loan interest, alimony paid, tuition & fees, Health Savings Accounts, one-half of the self-employment tax, self-employed health insurance, self-employed SEP, SIMPLE and qualified plans, and penalties on the early withdrawal of saving, minus other income. The Public Use File does not report other income. To calculate other income, we took the difference between reported AGI and the signed sum of the reported components of AGI. The difference includes moving expenses, foreign earned income, and net operating losses from previous years as well as errors due to censoring and rounding. Returns of dependents are excluded from the analysis. 41 See the notes to Table 7 for decile breakpoints.

32

expenditure for the state and local income tax deduction. The first column shows the distribution

of those taxpayers that claim an itemized deduction for state and local income taxes across

income deciles. The second column reports the average deduction. The results are not surprising

given what is known about the distribution of this expenditure from the JCT distribution tables,

for example. The benefits are highly concentrated at the top of the distribution (as is tax liability

and the group of taxpayers that itemize their returns).

Our analysis both complements the JCT analysis by showing information by income

deciles (with detail on the top decile) and adds to the analysis by considering how the distribution

changes under alternate AMT scenarios. Without the AMT, the benefit of this itemized deduction

is skewed even more to the top of the distribution. Without an AMT patch, on the other hand, the

benefits are somewhat less skewed across the income deciles. However, the benefits do become

more concentrated within the top one percent to returns.

Table 9 shows the distribution of the special rates on dividends and capital gains. Given

the concentration of this income in the top decile, this distribution of benefits is not surprising.

Nevertheless, it is striking that 99 percent of the benefit goes to taxpayers in the top five percent

of the distribution. Moreover, the distribution of this tax expenditure is less progressive than the

others reported in this paper.

Table 10 shows the distribution of our hypothetical tax expenditure that bundles the top

thirteen tax expenditures in TAXSIM. Taxpayers across all deciles benefit from this group of tax

expenditures, with the earned income tax credit and child credit most important in the bottom

deciles. Table 11 shows how the distribution of the top expenditures is affected by these two

credits by removing them from the analysis. While only about one half of the top thirteen

expenditures in TAXSIM go to the top ten percent of tax units, once we eliminate the EITC and

child credit, almost 70 percent of the top expenditures are claimed by this group. The average

expenditure in each of the bottom five deciles falls significantly and the total benefit received by

the bottom half falls from 20.5 percent to 4.5 percent.

33

Burman et al. find that eliminating most major tax expenditures simultaneously would

make the tax system less progressive as tax expenditures constitute a larger percentage of lower-

income taxpayer’s AGI than higher-income taxpayers. The analysis in our paper generally

confirms these results, as can be seen by comparing the distribution of liability in Table 7 to, for

example, the distributions reported in Tables 10 and 11.

In theory, the distributional consequences of tax expenditures are not limited to income

concepts. For example, unlike almost all revenue estimates, JCT and OTA report tax

expenditures estimates by taxpayer type: corporate or individual. While this additional

information is useful, it can also be misleading. Many business sectors are dominated by pass-

through entities that report tax liability on individual income tax forms. This conflates individual

income tax data with business tax data and can thus be misused in certain contexts. Furthermore,

like other tax variables, there may be uneven allocation of tax expenditure benefits across

geographic space. Unfortunately, OTA and JCT do not report spatial analyses of tax provisions

and proposals.42

TAX EXPENDITURE REPORTING RECOMMENDATIONS

After completing a review of official tax expenditure reporting, as well as an analysis of

estimates of current and hypothetical tax expenditures using TAXSIM, it is natural to contemplate

tentative answers to the questions we have raised thus far. In this section we present tentative

answers in the form of recommendations for expenditure reporting reform. Before discussing

these reforms, we should first note a conclusion all analysts draw when working with tax

microdata: it is not as easy as it looks and it is easier to criticize than it is to implement change.

Furthermore, we recognize that adopting the following recommendations will entail more work

42 For a recent examples of the spatial estimation of housing-related tax expenditures, see Gyourko and Sinai (2004) and Dietz (2006). For recent information regarding the deductibility of state and local taxes, see CBO (2008).

34

for government economists. Nonetheless, we believe our recommendations could improve the

quality and usefulness of tax expenditure reporting.

Should tax expenditure reports include behavior found in revenue estimates?

No. We believe that tax expenditures should be reported with no projected change in

economic behavior to capture how the tax system induces behavior. The original intent was to

present an estimate of government resources dedicated to a particular tax policy provision. This

intent assumes that behavior will be modified by the tax policy and this induced behavior should

be reflected in the revenue estimate.

Should revenue estimates accompany tax expenditure estimates?

Yes, at least for major tax expenditures that affect economic behavior. Too many

observers use tax expenditure estimates incorrectly as revenue estimates. While it would increase

the workload of the government economists who estimate tax expenditures, dual reporting would

provide useful additional information, as well as indicate tax expenditures inducing considerable

economic activity. For example, as discussed in the previous section, the tax expenditure for the

mortgage interest deduction is clearly not equivalent to the revenue estimate of repealing the

mortgage interest deduction. The revenue estimate would be considerably smaller as taxpayers

modify their behavior in response to the policy change.

Implementing this recommendation raises a number of interesting issues. First, how

would government agencies determine which expenditures should have accompanying revenue

estimates? Second, revenue estimators would need to be provided with details on how the

proposal to remove the expenditure in question would be structured. The JCT could resolve these

thorny issues by including revenue estimates for the revenue options presented in the bi-annual

CBO Budget Options report. Since the JCT economists prepare the estimates for the CBO report,

the extra revenue estimating work required would be minimal. This type of presentation is likely

35

to have a large impact on the public debate of popular tax expenditures by making a clear

distinction between revenue and tax expenditure estimates. Indeed, the staff of the JCT has

recently indicated that “where helpful and feasible” the staff will report accompanying revenue

estimates. (see JCT 2008)43

Should tax expenditures assume present-law sunsets?

Yes, but tax expenditures should be reported for permanently extended variants in an

appendix for major tax expenditures. We have demonstrated the significant effects sunsets of tax

policies can have on tax expenditures. As we noted, in addition to the direct effect of expiring

provisions, the expiration of the EGTRRA and JGTTRA yield changes for all other tax

expenditures. For such major sunsets, additional reporting should be provided in an appendix.

Should negative tax expenditures be reported?

Yes. Tax expenditures are reported because they are deviations from a comprehensive

income tax. Using the same principles, overly-burdensome aspects of tax laws should

correspondingly be reported. If economists are comfortable discussing negative income tax rates,

there is no reason not to report negative tax expenditures. Moreover, the accurate estimation or

classifying of some positive tax expenditures requires a corresponding reporting of certain

negative tax expenditures.44 Interestingly, the staff of the JCT has announced that it will report

negative expenditures in the future (JCT 2008).

How should the AMT be treated for tax expenditure estimation?

43 See JCX-37-08, page 83. 44 An example would be the entity-level taxation of business income and properly estimating the tax expenditure associated with the double taxation of corporate income.

36

An appendix should report estimates for tax expenditures with no AMT and under AMT

patch projections.45 We have demonstrated the dramatic effect that the AMT can have on

particular tax expenditure estimates. For example in the second section, we demonstrated that the

tax expenditure estimate for the state and local income tax deduction for 2007 increases from $26

billion to $33 billion under a hypothetical repeal of the AMT and falls to $16 billion with no

AMT-patch. Given such striking changes in the estimates, changes that are not substantively

related to changes in the rules regarding the income tax deduction or even forecasts of local

government tax polices, we believe that an AMT appendix should be available for estimates of

major tax expenditures illustrating interactions with the AMT. Further, there are other major

interactive features of the tax code that limit the scale of other tax benefits, such as the Pease

rules and the personal exemption phaseout rules. Each rule increases tax liability for certain high-

income taxpayers. An estimate of how such limiting rules affect major tax expenditures should

be reported for illustrative purposes.

Should additional tax expenditures be reported about “normal” aspects of the income tax?

Yes. As noted previously, many aspects of what is normal were put into practice without

proper consideration of the economics of these rules. Moreover, simply because a part of the

income tax has overwhelming public and expert support, such as the progressive system of rates,

does not mean it should not be quantified for analysis purposes. Reporting tax expenditures for

such basic elements of the tax system as the progressive system of rates, the standard deduction

and the personal exemption would place useful information in the public domain, particularly

when rate changes are considered. The estimates provided in the second section indicate the

prominent role that such features play in the tax code.

45 Analysts would be required to decide which AMT options to include. Theoretically there are many different AMT reforms that could be considered. However for information purposes including the “no AMT” and “no patch” seems most sensible.

37

Should distributional analysis be included in tax expenditure reports?

The JCT inclusion of distributional analysis for major tax expenditures provides

important information to its audience. We believe that this presentation should continue and

should be part of the Administration Budget presentation. That said, the process of deciding how

to best present distributional tables and what provisions to distribute is fraught with controversy.

There is no one “correct” way to present a distributional table. Whether expanded, cash, or

economic income should be used to classify taxpayers, what unit of analysis to use, whether to

present by deciles or income classes, what information to present, and which time periods to

show, for example, are some of many questions analysts confront when constructing distribution

tables. Distributions that show groups of expenditures, like those in Burman et. al., are useful to

the policy debate and could easily be put forward by JCT and OTA. We encourage both agencies

(as well as CBO) to study options for the presentation of distributions of tax expenditure analysis

as part of the annual tax expenditure reports.46

Should tax expenditures be reported by provisions or conceptually-linked categories?

We believe that it would be useful to report group tax expenditures for such conceptually-

linked areas as housing and health. As we demonstrated in the third section, the sums of

individual tax expenditures can be very different from the tax expenditure for a bundle of tax

provisions. Given the tendency of analysts to incorrectly sum individual tax expenditures to

provide estimates of tax expenditures for such areas as housing and health, we believe group

estimates would be an important addition. We further believe it would not be particularly useful

to group to tax expenditures by tax function (such as all credits or all deductions) given the lack

of linkages between such groupings.

46 See Joint Committee on Taxation (1993) for an extensive discussion of the methodological issues involved in producing distributional analysis. See Cronin (1999) for a discussion of the OTA distribution methodology.

38

We note caution should be used if tax expenditures are reported by such purposes as

“policy” or “subsidy.” If such classifications are to be made, and we are agnostic on this point,

the arguments for and against such policies should be presented, as is now done in many JCT

documents. For example, if a particular tax expenditure is intended to promote a positive

externality, it should be noted as such. Too often the tax expenditure list is consider a “hit list” of

possible future revenue raisers, as if the reporting of a particular provision as a tax expenditure is

a de facto condemnation of such a provision as a loophole in need of closing. As noted in the

beginning of this paper, the JCT recently indicated that it will employ similar categories to

classify tax expenditures and may report on the economic and tax policy consequences of selected

tax expenditures.

Should tax expenditures be grouped with fundamental tax reform considerations in mind?

Maybe. Given that the major candidates for fundamental tax reform are broadly known,

we believe it would be useful to report tax expenditure estimates in a possible “tax reform”

appendix that are grouped by fundamental tax reform proposals. For example, the set of itemized

deductions could be estimated simultaneously as a means of demonstrating the scale of such

reforms as a flat tax. Other such groups could include savings/investment incentives, education

incentives and certain international tax policy considerations (such as a territorial system).

Although it is natural to link tax expenditure analysis with tax reform, simply showing the tax

expenditure estimate associated with any reform may not be particularly useful or informative.

Consider a reform that eliminates all itemized deductions. Simply reporting the revenue raised is

misleading since it is likely that those revenues would be returned to taxpayers through lower

rates and/or restructured tax expenditures (for example, refundable credits). The reform would

39

have revenue, distributional, and efficiency consequences that require a much fuller analysis than

can be provided through tax expenditure estimation.47

How should tax expenditures with overlapping qualifying criteria be estimated?

Tax expenditure estimates should include “tax form behavior” or substitution among tax

provisions that have overlapping qualifying criteria, such as various investment and savings

incentives. For example, we demonstrated such a possibility among various education credits,

including the effect adopting various assumptions has on the respective tax expenditures. As no

economic behavior has changed in these situations, tax expenditures should reflect such

substitutions on the tax form.

CONCLUSION

Despite the challenges associated with tax expenditure classification and estimation, the

annual reporting of tax expenditure estimates is an important source of policy analysis for

economists and tax law observers. We have reviewed the evolution of the tax expenditure

reporting process and identified examples of inconsistencies between present and theoretically

ideal practice. Using NBER’s TAXSIM, we have demonstrated the complications produced by

the AMT, the standard deduction, the grouping of tax expenditures, expiring provisions, and

interactive effects in the estimation process. Mindful of these issues, we have suggested

incremental and implementable reforms to improve the reporting and use of tax expenditure

estimates. These include corresponding revenue estimates for major tax expenditures, negative

tax expenditure reporting, and establishing an appendix for tax expenditure estimates of

permanent versions of expiring provisions and AMT interaction effects, among others. These

proposed changes to existing tax expenditure estimation process will increase the value of tax

expenditure reporting for academics, policymakers, and others in the tax analysis community. 47 We thank Eric Toder for raising this point.

40

REFERENCES

Ackerman, Jonathan Z. and Rosanne Altshuler (2006). “Constrained Tax Reform: How Political and Economic Constraints Affect the Formation of Tax Policy Proposals.” National Tax Journal, 59(1): 165-187.

Andrews, William D. (1972). “Personal Deductions in an Ideal Income Tax.” Harvard Law

Review 86. Andrews, William D. (1984). “A Source of Inspiration.” Harvard Law Review 98(2): 332-334. Barthhold, Thomas A., Thomas Koerner, and John F. Navratil (1998). “Effective Marginal Tax

Rates under the Federal Individual Income Tax: Death by One Thousand Pin Pricks?” National Tax Journal 51(3): 553-64.

Bittker, Boris I. 1969. "Accounting for Federal 'Tax Subsidies' in the Budget." National Tax

Journal 22(2):244-61. Burman, Leonard (2003). “Is the Tax Expenditure Concept Still Relevant?” National Tax Journal

56(3):613-628. Burman, Leonard, Christopher Geissler, and Eric Toder (2008). “The Growth, Distribution and

Opportunity Cost of Individual Tax Expenditures.” Tax Policy Center Discussion Paper. Forthcoming.

Carroll, Robert, David Joulfaian, and James Mackie (2008). “Income versus Consumption Tax

Baselines for Tax Expenditures,” mimeo. Century Foundation (2002). Bad Breaks All Around: The Report of the Century Foundation

Working Group on Tax Expenditures. New York: The Century Foundation Press. Congressional Budget Office (2008). The Deductibility of State and Local Taxes. February. Congressional Budget Office (2007). Budget Options. February. Cronin, Julie-Anne (1999). “U.S. Tresury Distributional Analysis Methodology,” Office of Tax

Analysis Working Paper 85, Washington, DC. Davie, Bruce (1994). “Tax Expenditures in the Federal Excise Tax System.” National Tax

Journal 47(1): 39-62. Dietz, Robert (2006). “Local Use of the Mortgage Interest and Real Estate Tax Deductions.”

Housing Economics Online. Washington, D.C.: National Association of Home Builders, June.

Feenberg, Daniel R. and James M. Poterba (2004). “The Alternative Minimum Tax and Effective

Marginal Tax Rates,” National Tax Journal 57(2): 407-27. Fiekowsky, Seymour (1980). “The Relation of Tax Expenditures to the Distribution of the "Fiscal

Burden.” Canadian Taxation 2 (Winter): 211-219.

41

Gyourko, Joseph and Todd Sinai (2004). “The (Un)Changing Geographic Distribution of Housing Tax Benefits, 1980-2000.” In J. Poterba, ed., Tax Policy and the Economy 18, 165-208.

Hungerford, Thomas L. (2006). “Tax Expenditures: Trends and Critiques.” Library of Congress,

Congressional Research Service Report. RL33641, September. Joint Committee on Taxation (2008). “A Reconsideration of Tax Expenditure Analysis,” JCS-37-

08. Joint Committee on Taxation (2007). “Estimates of Federal Tax Expenditures For Fiscal Years

2007-2011,” JCS-3-07. Joint Committee on Taxation (2007). “Present Law and Background Relating to the Individual

Alternative Minimum Tax,” JCX-38-07. Joint Committee on Taxation (2005). “Options to Improve Tax Compliance and Reform Tax

Expenditures,” JCS-02-05. Joint Committee on Taxation (1993). “Methodology and Issues in Measuring Changes in

the Distribution of Tax Burdens,” JCS-7-93. Leiserson, Greg (2007). “The 15 Percent Rate on Capital Gains: A Casualty of the Alternative

Minimum Tax,” http://www.urban.org/url.cfm?ID=901052. Polackova, Hana Brixi, Christian M.A. Valendue, and Zhicheng Swift, eds. (2004). Tax

Expenditure: Shedding Light on Government Spending through Taxation. Washington, DC: World Bank.

Poterba, James and Todd Sinai (2008). Revenue Cost and Incentive Effects of Tax Expenditures

on Owner-Occupied Housing, mimeo. Neubig, Thomas S. and David Joulfaian (1988). “The Tax Expenditure Budget Before and After

the Tax Reform Act of 1986,” U.S. Treasury Department OTA Paper 60, October. President’s Advisory Panel on Federal Tax Reform (2005). Simple, Fair, and Pro-Growth:

Proposals to Fix America’s Tax System. Washington, DC: November. Shaviro, Daniel (2003). “Rethinking Tax Expenditures and Fiscal Language.” Mimeo. Steuerle, C. Eugene (2004). Contemporary U.S. Tax Policy. Washington, DC: Urban Institute

Press, 43-44. Surrey, Stanley S. (1967). The U.S. Income Tax System – the Need for a Full Accounting.

Speech before the Money Marketers, November 15. Surrey, Stanley S. (1973), Pathways to Tax Reform. Cambridge, Mass.: Harvard University Press. Surrey, Stanley S. and Paul R. McDaniel (1985). Tax Expenditures. Cambridge, Mass: Harvard

University Press.

42

43

Thuronyi, Victor (1988). “Tax Expenditures: A Reassessment,” Duke Law Journal 1988(6): 1155-1206.

Toder, Eric (1998). “The Changing Composition of Tax Incentives: 1980-1999.” National Tax

Association – Proceedings of the 91st Annual Conference, 411-420. Toder, Eric (2002). "Evaluating Tax Incentives as a Tool for Social and Economic Policy." In

Bad Breaks All Around: Report of the Century Foundation Working Group on Tax Expenditures. New York: Century Foundation Press.

Toder, Eric (2005). “Tax Expenditures and Tax Reform: Issues and Analysis.” National Tax

Association – Proceedings of the 98th Annual Conference, 472-479. U.S. General Accountability Office (2005). “Tax Expenditures Represent a Substantial Federal

Commitment and Need to Be Reexamined,” GAO-05-690. Washington, D.C.: September.

U.S. Office of Management and Budget (2008). Analytical Perspectives, Budget of the United

States Government, Fiscal Year 2009, Washington, DC: U.S. Government Printing Office.

Weiner, Joann (2008). Tax Expenditures and the Federal Budget, Tax Notes, February 4, 586-

592.

Table 1 A Comparison of Baselines

JCT Normal Tax Treasury Normal Tax Treasury Reference Tax

Individual tax • One personal exemption for each taxpayer and dependent • Standard deduction • Includes all cash transfer payments from the Government • Deductions for investment and employee business expenses • Prevailing tax rate schedule and tax brackets • Capital gains taxed upon realization • All employee compensation subject to tax currently

o Employee stock options taxed at regular rate when options exercised (with corresponding deduction for employees). Income is difference between purchase price of stock and the market price on the day the option exercised.

• All other income and transfers subject to tax o Social security income excluded only for the portion of retirement benefits that represent

a return of payroll taxes paid during working years; Medicare benefits excluded only for the portion of HI tax contributions; public assistance benefits (food stamps, Medicaid, public housing) subject to tax; gifts excluded

• Imputed income from owner-occupied homes excluded (but not classified as tax expenditure due to administrative necessity)

• Income tax levied on nominal not real gains in asset values (no indexing) • Foreign tax credit

JCT normal tax with the following exceptions • Includes prevailing rates on

capital gains on corporate equity and dividends (since 2005 Budget)

• Includes AMT and passive loss rules as part of the baseline

• Includes net imputed rental income

Treasury normal tax with the following exception • Gross income does not include

transfer payments

Business income taxation • Treatment of capital costs

o Cost recovery allowances more favorable than straight-line recovery o No indexing

• Accounting standards o Accrual method of accounting, standard of “economic performance” used to determine

whether liabilities are deductible, and general concept of matching income and expenses. Tax provisions that do not satisfy all three are viewed as tax expenditures.

• Prevailing carryback and carryforward periods for net operating losses • Top statutory rate on corporate income (no graduated rates) • Special tax rules for pass-through entities and nonprofit corporations exempting them from

corporate income tax • Controlled foreign corporations (CFCs) not considered separate entities from controlling

U.S. shareholders • Foreign tax credit

JCT normal tax with the following exceptions • Includes corporate AMT • Includes cash method of

accounting for certain businesses

• Uses economic depreciation in baseline

Treasury normal tax with the following exceptions • Includes prevailing graduated

corporate rates • Includes accelerated

depreciation • CFCs are considered separate

entities (except for tax haven activities)

Notes: See JCT (2007) and OMB (2008) for further details

44

Table 2 Tax Expenditure Estimates for a Hypothetical Taxpayer

Base case (1)

Eliminate state and

local deduction

(2)

Eliminate real estate

deduction (3)

Eliminate home

mortgage interest

deduction (4)

Eliminate charitable

contributions deduction

(5) Itemized deductions:

State and local income taxes $6,500 $0 $6,500 $6,500 $6,500Real estate taxes 3,500 3,500 0 3,500 3,500Home mortgage interest deduction 1,100 1,100 1,100 0 1,100Charitable contributions 2,000 2,000 2,000 2,000 0Sum of itemized deductions 13,100 6,600 9,600 12,000 11,100

Standard deduction 10,000 10,000 10,000 10,000 10,000Taxpayer itemizes? Yes No No Yes YesItemized deductions claimed by taxpayer 13,100 10,000 10,000 12,000 11,100Tax expenditure estimate for eliminated itemized deduction(s) 620 620 220 400

Tax expenditure if standard deduction raised to $12,000 220 220 220 220Tax expenditure if itemized deduction were an above the line deduction or an adjustment 1,300 700 220 400

Note: Assumes a flat marginal tax rate of 20 percent.

45

Figure 1Effect of AMT on Select Tax Expenditure Estimates, 2000-2007

0

10

20

30

40

50

60

70

80

90

100

2000 2001 2002 2003 2004 2005 2006 2007

Current Law Current Law with No AMT Current Law with No AMT Patch

Mortgage interest deduction

State and local income tax deduction

Property tax deduction

Notes: Authors' calculations using NBER TAXSIM. See text for further details.

Billions of dollars

46

Figure 2Effect of AMT on Selected Tax Expenditure Estimates, 2007

0

20

40

60

80

100

120

Mortgage interestdeduction

Property taxdeduction

State and localincome tax deduction

Special rates fordividends and capital

gains

Child tax credit Lifetime and HopeCredits

Bill

ions

of d

olla

rs

Current Law Current Law with No AMT Current Law with No AMT Patch

Notes: Authors' calculations using NBER TAXSIM. See text for further details.

47

Figure 3Tax Expenditure Estimate for Graduated Individual Marginal Tax Rates, 2005

0

50

100

150

200

250

300

350

10 percent 15 percent 25 percent 28 percent 33 percentMarginal Tax Rates

Bill

ions

of d

olla

rs

Notes: Authors' calculation from SOI data. See text for details.

48

Figure 4 Tax Expenditure Estimate for Mortgage Interest Deduction, 2000-2011

0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Billions of dollars

Notes: Authors' calculations using NBER TAXSIM. See text for further details.

Current LawPre-EGTRRA Law

Current Law w/ Flat Rate

49

Figure 5 Tax Expenditure Estimate for State and Local Income

Tax Deduction, 2000-2011

0

10

20

30

40

50

60

70

80

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Billions of dollars

Notes: Authors' calculations using NBER TAXSIM. See text for further details.

Pre-EGTRRA Law

Current Law w/ Flat Rate

Current Law

50

Figure 6Tax Expenditure Estimate for Property Tax Deduction, 2000-2011

0

5

10

15

20

25

30

35

40

45

50

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Billions of dollars

Notes: Authors' calculations using NBER TAXSIM. See text for further details.

Current Law

Pre-EGTRRA Law

Current Law w/ Flat Rate

51

Table 7 Distribution of Individual Income Tax Revenues under three AMT Scenarios

(in millions of dollars) Tax liability under 2004 law

Income decile Assuming 2004 AMT

Percentage of liability

Assuming no AMT

Percentage of liability

Assuming no AMT patch

Percentage of liability

Bottom 0 -0.8 -54 -0.8 1 -0.82nd -6,286 -1.6 -6,294 -1.6 -6,282 -1.53rd -12,671 -1.0 -12,684 -1.0 -12,671 -1.04th -8,114 0.0 -8,127 0.0 -8,114 0.05th -28 1.8 -33 1.8 -26 1.86th 14,445 3.3 14,405 3.4 14,460 3.37th 26,985 5.4 26,965 5.5 27,058 5.38th 43,264 8.0 43,211 8.2 43,440 7.99th 64,898 13.0 64,750 13.2 65,444 13.1Top 104,805 71.8 104,409 71.4 108,416 72.0

Detail on top decile Top 5% 484,370 60.0 471,885 59.5 494,337 59.8Top 1% 311,965 38.7 305,602 38.5 312,482 37.8Total 806,894 100.0 792,973 100.0 826,335 100.0

Notes: Calculations from NBER TAXSIM using 2004 Statistics of Income Public Use File. The income concept used to place tax returns into income classes is adjusted gross income (AGI) minus state and local tax refunds, plus tax-exempt interest, non-taxable Social Security benefits, deductions for IRA contributions, student loan interest, alimony paid, tuition & fees, Health Savings Accounts, one-half of the self-employment tax, self-employed health insurance, self-employed SEP, SIMPLE and qualified plans, and penalties on the early withdrawal of saving, minus other income. The Public Use File does not report other income. To calculate other income, we took the difference between reported AGI and the signed sum of the reported components of AGI. The difference includes moving expenses, foreign earned income, and net operating losses from previous years as well as errors due to censoring and rounding. Returns of dependents are excluded from the analysis. Returns with negative income are excluded from lowest income group but included in totals. Decile breakpoints are, respectively, $8,425, $14,618, $20,370, $27,310, $35,280, $44,876, $57,070, $73,865, and $104,687 ($145,801 for top 5% and $360,419 for top 1%).

52

Table 8

Distribution of the Tax Expenditure for the State and Local Income Tax Deduction under three AMT Scenarios, 2004 Money amounts in millions of dollars, returns in thousands

Under 2004 law Under 2004 law with no AMT Under 2004 law with no AMT patch

Income decile

Percentage of all

returns claiming

deduction Average

deduction

Percentage of total

deduction claimed

Percentage of all

returns claiming

deduction Average

deduction

Percentage of total

deduction claimed

Percentage of all

returns claiming

deduction Average

deduction

Percentage of total

deduction claimed

Bottom 0.0 19 0.0 0.0 19 0.0 0.0 19 0.0 2nd 0.3 56 0.0 0.3 56 0.0 0.4 56 0.0 3rd 1.0 86 0.1 1.0 86 0.1 1.1 86 0.1 4th 2.6 125 0.2 2.5 125 0.2 2.9 125 0.3 5th 5.7 172 0.8 5.6 172 0.6 6.4 172 0.9 6th 9.4 269 1.9 9.2 269 1.6 10.4 269 2.4 7th 12.7 395 3.8 12.5 395 3.2 14.0 391 4.7 8th 17.2 490 6.4 16.9 492 5.4 18.8 488 7.9 9th 23.7 803 14.5 23.4 813 12.4 23.8 794 16.2 Top 27.4 3,450 72.2 28.5 4,117 76.4 22.2 3,546 67.5

Detail on top decile

Top 5% 13.7 5,526 57.9 14.8 6,604 63.8 10.8 6,177 57.2 Top 1% 2.8 18,429 38.7 3.0 20,166 39.8 3.0 18,757 47.7 Total 100.0 1,309 100.0 100.0 1,535 100.0 100.0 1,167 100.0

See notes to table 7.

53

Table 9

Distribution of Hypothetical Tax Expenditure for Special Rates on Dividends and Capital Gains under three AMT Scenarios Money amounts in millions of dollars, returns in thousands

Under 2004 law Under 2004 law with no AMT Under 2004 law with no AMT patch

Income decile

Percentage of all

returns claiming

deduction Average

deduction

Percentage of total

deduction claimed

Percentage of all

returns claiming

deduction Average

deduction

Percentage of total

deduction claimed

Percentage of all

returns claiming

deduction Average

deduction

Percentage of total

deduction claimed

Bottom 0.0 0 0.0 0.0 0 0.0 0.0 0 0.0 2nd 0.0 0 0.0 0.0 0 0.0 0.0 0 0.0 3rd 0.0 46 0.0 0.0 46 0.0 0.0 46 0.0 4th 0.0 244 0.0 0.0 244 0.0 0.0 244 0.0 5th 0.1 56 0.0 0.1 56 0.0 0.1 56 0.0 6th 0.7 596 0.0 0.7 603 0.0 0.7 603 0.0 7th 1.3 685 0.0 1.4 685 0.0 1.3 686 0.0 8th 3.8 868 0.1 3.7 845 0.1 4.0 981 0.1 9th 10.8 1,221 0.4 10.8 1,183 0.4 10.8 1,369 0.5 Top 83.2 39,777 99.5 83.3 41,160 99.5 83.0 39,777 99.4

Detail on top decile

Top 5% 72.1 45,549 98.7 72.3 47,121 98.9 71.9 45,508 98.5 Top 1% 27.2 109,377 89.5 27.3 113,418 90.0 27.1 109,346 89.3 Total 100.0 33,273 100.0 100.0 34,452 100.0 100.0 33,225 100.0

See notes to table 7.

54

Table 10

Distribution of Hypothetical Tax Expenditure Combining the Top 13 TAXSIM Tax Expenditures under three AMT Scenarios Money amounts in millions of dollars, returns in thousands

Under 2004 law Under 2004 law with no AMT Under 2004 law with no AMT patch

Income decile

Percentage of all

returns claiming Average

Percentage of total claimed

Percentage of all

returns claiming Average

Percentage of total claimed

Percentage of all

returns claiming Average

Percentage of total claimed

Bottom 6.8 1,087 2.0 6.8 1,087 1.9 6.8 1,087 2.02nd 7.9 2,185 4.7 7.9 2,184 4.6 7.9 2,185 4.73rd 7.3 2,442 4.8 7.3 2,442 4.7 7.3 2,442 4.84th 8.5 2,119 4.9 8.5 2,119 4.8 8.5 2,119 4.95th 9.0 1,696 4.1 9.0 1,696 4.0 9.0 1,696 4.16th 10.3 1,720 4.8 10.3 1,720 4.6 10.3 1,722 4.87th 11.2 1,994 6.0 11.2 1,994 5.9 11.2 1,997 6.08th 12.2 2,218 7.3 12.2 2,218 7.1 12.2 2,240 7.49th 12.9 2,942 10.3 12.9 2,948 10.0 12.9 3,155 11.0Top 13.8 13,762 51.1 13.8 14,484 52.4 13.8 13,609 50.4

Detail on top decile

Top 5% 7.0 23,079 43.4 7.0 24,472 44.8 7.0 22,824 42.8Top 1% 1.4 81,163 30.7 1.4 85,142 31.3 1.4 80,804 30.5Total 100.0 3,705 100.0 100.0 3,805 100.0 100.0 3,715 100.0

See notes to table 7.

55

Table 11

Distribution of Hypothetical Tax Expenditure Combining 11 TAXSIM Tax Expenditures under three AMT Scenarios Money amounts in millions of dollars, returns in thousands

Under 2004 law Under 2004 law with no AMT Under 2004 law with no AMT patch

Income decile

Percentage of all returns

claiming Average

Percentage of total claimed

Percentage of all

returns claiming Average

Percentage of total claimed

Percentage of all

returns claiming Average

Percentage of total claimed

Bottom 0.1 84 0.0 0.1 88 0.0 0.1 84 0.02nd 2.2 272 0.2 2.2 272 0.2 2.2 272 0.23rd 4.0 552 0.6 4.0 552 0.6 4.0 552 0.64th 6.4 762 1.3 6.4 762 1.2 6.4 762 1.35th 8.9 1,026 2.4 8.9 1,025 2.3 8.9 1,026 2.46th 11.8 1,275 3.9 11.8 1,275 3.8 11.8 1,279 3.97th 13.8 1,557 5.6 13.8 1,557 5.4 13.8 1,562 5.68th 15.6 1,656 6.7 15.6 1,655 6.5 15.6 1,682 6.89th 17.6 2,322 10.6 17.6 2,329 10.2 17.6 2,548 11.6Top 19.5 13,624 68.9 19.5 14,355 70.0 19.5 13,478 67.8

Detail on top decile

Top 5% 10.0 23,072 59.5 10.0 24,468 60.9 10.0 22,818 58.7Top 1% 2.0 81,175 42.2 2.0 85,176 42.6 2.0 80,817 41.9Total 100.0 3,860 100.0 100.0 4,003 100.0 100.0 3,876 100.0

See notes to table 7.

56

Appendix Figure 1Distribution of Tax Expenditure for Marginal Rate Brackets, 2005

0

20

40

60

80

100

120

140

160

180

200

Under$10,000

$10,000 to$20,000

$20,000 to$30,000

$30,000 to$40,000

$40,000 to$50,000

$50,000 to$75,000

$75,000 to$100,000

$100,000 to$200,000

$200,000 to$500,000

$500,000 to$1,000,000

over$1,000,000

Income Class

Bill

ions

of D

olla

rs